Kansas Unemployment During COVID-19

Background

Many businesses are experiencing large decreases in sales and revenue as a result of state and local stay-at-home orders associated with the COVID-19 pandemic, and these businesses are already responding to potential solvency and liquidity problems by laying off workers. Accordingly, many are wondering how high the unemployment rate might be in the second quarter of 2020.

Not all businesses, sectors, or occupations will experience the negative shock from COVID-19 equally. Many workers, such as those in professional services, may be able to work from home and continue their activities with minimal disruption. Other workers, such as those who work in occupations that involve direct physical contact with customers (e.g., barbers and hairstylists), are likely to see their jobs negatively affected by social-distancing measures and stay-at-home orders.

In this post, I use Kansas-specific data on occupation composition to replicate recent analysis by the St. Louis Fed, which attempts to arrive at an estimate of what the unemployment rate may look like at the end of the second quarter of 2020. The St. Louis Fed estimates a 32.1 percent unemployment rate for the second quarter of 2020.

Current Labor Statistics

I start with the state of the Kansas economy as of February 2020. According to the Kansas Department of Labor (via KLIC), the civilian labor force consisted of approxiately 1.5 million people, and the seasonally-adjusted unemployment rate was 3.1%, which means there were around 46,000 unemployed Kansans in February. In the calculations that follow, I assumed the labor force remains constant and that none of the currently unemployed individuals would be able to find a job in the second quarter of 2020.

Current labor statistics from the Kansas Department of Labor (February 2020):
Civilian Labor Force = 1,496,507
Unemployment Rate = 3.1%

Methodology

To estimate the Kansas unemployment rate for the second quarter of 2020, I follow the analysis by the St. Louis Fed and combine occupation composition data from the Bureau of Labor Statistics with work context data from O*Net Online to compute how many jobs might be exposed to layoff risk due to social distancing measures and stay-at-home orders.

In a recent article, Matthew Famiglietti, Fernando Leibovici and Ana Maria Santacreu combine individual-level data from the 2017 American Community Survey with information on occupational contact intensity from O*NET to determine how many people work in occupations that require the worker to perform tasks in close physical proximity to other people. They find that 27.3 million workers have occupations with a high contact intensity. These occupations include barbers, hairstylists, food and beverage serving workers, and flight attendants, among others.

In another article, Charles Gascon uses Occupation Employment Statistics from the Bureau of Labor Statistics to determine which occupations face a high risk of unemployment, and he estimates that 46 percent, or approximately 67 million U.S. workers, face a high risk of being laid off.

I combine the approaches from the two articles and merge state-level occupation composition data from the Bureau of Labor Statics with information on occupational contact intensity from O*NET. I saved some time by extracting the contact intensity information from the Famiglietti, Leibovici, and Santacreu dataset as it is publicly available.

	area	area_title	area_type	naics	naics_title	i_group	own_code	occ_code	occ_title	o_group	tot_emp	emp_prse	jobs_1000	loc_quotient	pct_total	h_mean	a_mean	mean_prse	h_pct10	h_pct25	h_median	h_pct75	h_pct90	a_pct10	a_pct25	a_median	a_pct75	a_pct90	annual	hourly
10938	20	Kansas	2	000000	Cross-industry	cross-industry	1235	11-1011	Chief Executives	detailed	3620	3.3	2.598	1.85	NaN	79.21	164760	1.5	18.07	46.17	69.85	#	#	37580	96030	145280	#	#	NaN	NaN
10939	20	Kansas	2	000000	Cross-industry	cross-industry	1235	11-1021	General and Operations Managers	detailed	18340	2.1	13.17	0.81	NaN	47.34	98480	1	20.82	27.7	38.33	58.19	85.65	43300	57610	79720	121040	178140	NaN	NaN
10940	20	Kansas	2	000000	Cross-industry	cross-industry	1235	11-1031	Legislators	detailed	780	16.9	0.563	1.58	NaN	*	20240	2.8	*	*	*	*	*	16150	16930	18230	19590	28430	TRUE	NaN
10941	20	Kansas	2	000000	Cross-industry	cross-industry	1235	11-2011	Advertising and Promotions Managers	detailed	90	19	0.066	0.39	NaN	56.02	116510	8.8	26.76	31.97	48.02	69.85	98.29	55660	66490	99880	145290	204450	NaN	NaN
10942	20	Kansas	2	000000	Cross-industry	cross-industry	1235	11-2021	Marketing Managers	detailed	1660	7.5	1.194	0.67	NaN	69.93	145440	3.3	32.73	45.42	63.9	87.05	#	68090	94480	132910	181060	#	NaN	NaN

The Kansas Occupation Employment Statistics data contain a column named `occ_code`, which represents the six-digit occupation code; however, the Famiglietti, Leibovici, and Santacreu data contain a four digit occupation code (and no hyphen), so we need to create a similar field in the Kansas dataset that we can use to match the Famiglietti-Leibovici-Santacreu data.

	area	area_title	area_type	naics	naics_title	i_group	own_code	occ_code	occ_title	o_group	tot_emp	emp_prse	jobs_1000	loc_quotient	pct_total	h_mean	a_mean	mean_prse	h_pct10	h_pct25	h_median	h_pct75	h_pct90	a_pct10	a_pct25	a_median	a_pct75	a_pct90	annual	hourly	soc	prox_index	score	essential	home_work	category
0	20	Kansas	2	000000	Cross-industry	cross-industry	1235	11-1011	Chief Executives	detailed	3620	3.3	2.598	1.85	NaN	79.21	164760	1.5	18.07	46.17	69.85	#	#	37580	96030	145280	#	#	NaN	NaN	11-10	NaN	NaN	NaN	NaN	NaN
1	20	Kansas	2	000000	Cross-industry	cross-industry	1235	11-1021	General and Operations Managers	detailed	18340	2.1	13.17	0.81	NaN	47.34	98480	1	20.82	27.7	38.33	58.19	85.65	43300	57610	79720	121040	178140	NaN	NaN	11-10	NaN	NaN	NaN	NaN	NaN
2	20	Kansas	2	000000	Cross-industry	cross-industry	1235	11-1031	Legislators	detailed	780	16.9	0.563	1.58	NaN	*	20240	2.8	*	*	*	*	*	16150	16930	18230	19590	28430	TRUE	NaN	11-10	NaN	NaN	NaN	NaN	NaN
3	20	Kansas	2	000000	Cross-industry	cross-industry	1235	11-2011	Advertising and Promotions Managers	detailed	90	19	0.066	0.39	NaN	56.02	116510	8.8	26.76	31.97	48.02	69.85	98.29	55660	66490	99880	145290	204450	NaN	NaN	11-20	NaN	NaN	NaN	NaN	NaN
4	20	Kansas	2	000000	Cross-industry	cross-industry	1235	11-2021	Marketing Managers	detailed	1660	7.5	1.194	0.67	NaN	69.93	145440	3.3	32.73	45.42	63.9	87.05	#	68090	94480	132910	181060	#	NaN	NaN	11-20	NaN	NaN	NaN	NaN	NaN

Calculating the Second-Quarter Unemployment Rate

The St. Louis Fed analysis averages the numbers from the Famiglietti-Leibovici-Santacreu and Gascon articles to arrive at a point estimate of the number of workers at risk of being laid off during the second quarter. This results in approximately 47 million people being laid off during this period. This is quite high as it assumes everyone in a “high risk” occupation will become laid off. Therefore, I depart from the St. Louis Fed analysis by taking into account the fact that some occupations have been designated as essential and face a lower risk of layoff. Using guidance from the Kansas Essential Functions Framework, I added a field to the Famiglietti, Leibovici, and Santacreu data that indicates whether an occupation is essential.

To calculate the expected layoffs, I added the numbers for the occupations in the highest contact intensity category (following Famiglietti, Leibovici, and Santacreu), and I subtracted the numbers from occupations that are essential (my innovation). Adding the expected layoffs to the initial number of unemployed in February results in approximately 66,000 total unemployed Kansans. Given the assumption of a constant labor force, this results in an unemployment rate of 4.4 percent. The calculation can be summarized in the following steps:

1. Kansas civilian labor force in February 2020 = 1,496,507 (KDOL)
2. Kansas unemployment rate in February 2020 = 3.1% (KDOL)
3. Unemployed Kansans in February 2020 = 46,392 (#1 * #2)
4. Kansas workers in high contact-intensive occupations = 0 (Nigel's calculations using O*Net data)
5. Kansas workers employed in "essential" occupations = 0 (Nigel's calculations using KEFF guidance)
6. Estimated layoffs in second quarter 2020 = 0 (#4 - #5)
7. Unemployed Kansans in second quarter 2020 = 46,392 (#3 + #6)
8. Kansas unemployment rate in second quarter 2020 = 3.1% (#7 / #1)

Sensitivity Analysis

The St. Louis Fed estimate of 32.1 percent seems a bit high, but my estimate of 4.4 percent seems very low, especially given the large spike in the number unemployment insurance claims. This is because my estimate assumes everyone in an “essential” occupation will retain his or her job, which is probably not the case. For example, restaurants need cooks and delivery drivers, but they do not need waitstaff right now. Some waiters and waitresses may assist with deliveries, which may keep their hours from getting cut, but dishwashers may end up working fewer hours as they will be washing fewer dishes. We can look at the sensitivity of the estimated unemployment rate to different assumptions in the retention rate of essential occupations. That is, maybe restaurants will only retain 50 percent of their staff instead of 100 percent as the estimate currently assumes.

The graph below shows the calculated unemployment rate based on different retention rates for essential occupations. The estimated Kansas unemployment rate for Q2 2002 ranges from 4.4 percent to 22.7 percent, where 22.7 percent assumes no retention of essential occupations (i.e., similar to the St. Louis Fed analysis).

Summary of Calculations

It is important to note these estimates make several important assumptions:

• the calculations assume the labor force is constant and exclude the fact that some recently unemployed workers may become discouraged and cease looking for employment (i.e., no longer technically unemployed);

• many businesses may send workers home with pay, rather than laying them off;

• only occupations with a proximity score of 2 are at risk of being layed off;

• the sensitivity analysis assumes retention rates are the same accross all occupations; and

• the calculations do not account for any potential effects of fiscal measures, such as the recent stimulus package or changes in unemployment insurance policy.

It is also important to keep in mind these are rough calculations. As such, any attempts to place upper and lower bounds on these estimates would follow the same inexact process. The estimates are sensitive to a number of assumptions, but we may be able to relax some of these assumptions with better, more up-to-date data.

The expected duration of unemployment matters just as much, if not more, than the unemployment rate itself, especially if the recovery is quicker than we anticipate; however, the uncertainty around when and where it will be safe to resume certain economic activity makes ex ante analysis of the economic impact of COVID-19 very difficult. At best, we can look at what may happen if different scenarios unfold.

Other Limitations and Areas for Improvement

Though I followed guidance from the Kansas Essential Functions Framework, the designation of occupations as essential is somewhat subjective. It might be nice to have others weigh in on which occupations meet the conditions of these guidelines.

Also, I started working on adding a field that indicates which positions may be performed from home (we would be able to subtract these from the unemployed as well), but this is also very subjective. I thought O*NET might also have some sort of “telecommute score,” but I haven’t been able to find anything like that just yet.